What this covers

Patrick of Invest Like the Best talks with Michael Mayer, who runs Bottomless, a company that sells a connected scale so staples like coffee can be tracked and reordered. The conversation starts with how the product was built: scales hand-assembled at about three hours each, then iterated in weekly small batches rather than perfected for a Kickstarter, a route Mayer thinks is overrated. It then moves to the economics behind it. The average grocery shopper buys only about 13 SKUs a week, so for staples the open question is when to ship. Making consumption legible to software is his answer. He also blames rising customer acquisition cost for most startup deaths after the early years.

The second half ranges well beyond commerce. Mayer describes spaced repetition as a way to store knowledge durably and imagines a social network that rewards lasting posts over ephemeral ones. He argues that coding is more accessible than people assume, that schooling's batch structure made sense only before self-education was feasible, and that topical news has a short information shelf life. Business strategy, he says, is already given away free in shareholder letters and books, which leaves execution and judgment as the scarce ingredients.

The more contested turns come late. He sketches a market cycle from artisanal to mass-produced to artisanal revival, citing beer and possibly cannabis. He separates equal opportunity from equal outcomes, arguing that policies chasing the latter drop people at both ends. He says a Harvard credential builds limited social capital because its holders are interchangeable. On social media, he says the aim should be to be interesting rather than true, and that popular ideas get misapplied because few people publicly resist them.

Sharpest takeaway

Mayer argues that durable advantage comes from making repeat purchases legible to software and from owning the customer relationship, while conventional advice on books, news, contrarian bets, and equality is mostly noise or misleading.

  • Subscriptions fail because consumption timing is unpredictable, and sensors that make stock levels legible fix the timing problem
  • Owning the direct customer relationship sets the default and gives upstream leverage over suppliers
  • Originality depends on a curated information diet, and most popular sources and contrarian ideas are already consensus

The argument · threads13 threads · 73 claims
0.85

Validating demand through small, frequent hardware builds beats polishing products before a large launch.

9 pointscentrality 5/5
  • Building and shipping many hardware iterations in small batches, rather than perfecting a product for a Kickstarter launch, produced a far better scale, because many hardware companies die after spending heavily on a crowdfunded product that has not been validated.

    “So, the iterative process got us to where we needed to be rather than just trying to dream up something and doing a Kickstarter.”

  • After the early years, rising customer acquisition costs are the main cause of startup death, because paid acquisition through Google and Facebook can consume roughly 40 cents of every venture dollar, as Chamath's annual letter suggests.

    “I think that's the number one cause of startup death after the early years. Just your acquisition costs blowing up.”

  • Investor validation is important because convincing smart people to part with their money, especially angels investing personal funds, is meaningful evidence that the vision is credible, contrary to the common advice to ignore it.

    “people say that you shouldn't care about investor validation, but I think that's bullshit. Investor validation is important.”

  • Platform builders cannot foresee use cases, but relying on emergent behavior as strategy is overrated; founders should emulate deliberate business plans like Amazon's and Bezos's rather than iterating toward an epiphany.

    “In general, I think, business strategy, depending on emergent behavior is overrated right now.”

  • Asking customers what they would need to be paid to cancel revealed that some valued Bottomless at around $1,000, because it saved them trips and time that they projected over years, which Mayer treats as a measure of consumer surplus.

    “When people told me $1,000, and I said, 'No.'”

  • Early pilots with small samples, such as 15 customers, produce high anxiety because losing one customer can feel like failure, even though small samples cannot support strong conclusions.

    “Small sample sizes can create a lot of anxiety.”

  • Iterating on the scale raised its reliability from working about 70% of the time to about 95%, and it became easy enough to assemble that the design could be handed to an engineer for PCB manufacturing.

    “It works 95% of the time instead of 70% of the time.”

  • The most authentic pushback against Bottomless is whether people want all their stuff monitored by sensors and shipped to them, and skepticism about this audacious long-term vision is rational because no one knows the answer yet.

    “I think skepticism around that is probably the most authentic pushback because, nobody knows.”

  • Early Bottomless scales were hand-built at about three hours each and unpainted without customer complaint, and the team iterated by building units weekly rather than perfecting a product and running a Kickstarter, which many hardware companies do and which Michael considers overrated.

    “We've learned so much making different versions of this thing. The version we have now is way better than what we had before.”

0.65

Giving hardware away and earning from recurring consumables is how Bottomless wins.

7 pointscentrality 4/5
  • The average grocery consumer buys only about 13 SKUs every week, meaning shoppers repeatedly buy the same products, so the retail information problem for most staples is when to ship rather than whether the customer wants the item.

    “I saw this study that the average grocery consumer buys only 13 SKUs every week.”

  • Subscriptions for repeatedly purchased goods fail not because people do not want the items but because consumption timing is unpredictable, so even consistent consumers would churn within three or four shipments on a fixed schedule.

    “even very consistent consumers for us, if we were to put them on a set subscription schedule, they would get off very fast - within three or four shipments.”

  • Bottomless gives the scale away for free and makes money on the coffee in the subscription, because customers do not want to pay for reordering hardware and already have the setup they need.

    “we can give this away to people for free and make money off of the actual coffee in the subscription.”

  • The scale was designed to run for at least one year on battery because a monthly reorder cycle would make frequent recharging defeat the purpose of an automatic replenishment device.

    “So, we had a design goal of having the scales last for at least a year.”

  • Coffee roasters readily join Bottomless because they already sell online through Shopify, they already ship individual bags, and customers who over-subscribe complain about stale coffee, so the platform helps them deliver coffee at its freshest.

    “They see people complain saying, 'Hey, I have five bags, and I want a refund for four of them because they're all going stale.'”

  • Because customers receive a free scale and must return it to switch, the physical device creates SaaS-like stickiness that makes a future competitor unlikely to win the customer.

    “So there's this massive stickiness that you typically would see in SaaS companies that we have”

  • The company deliberately keeps the device minimal, with one button for WiFi sync, and does not plan to add features, because the strategy depends on giving hardware away and earning from consumables.

    “I don't think in the future we're going to add any bells and whistles. It's just one button to do the WiFi sync, and that's it.”

0.55

Transferable social capital and early talent signals matter more than replaceable credentials, regardless of pseudonymity.

4 pointscentrality 4/5
  • Social capital built through pseudonymous accounts is still one's own and can be transferred to a real identity when needed, so accumulating it only under one's own name is overrated.

    “So, it's somewhat overrated to accumulate social capital only under your own name.”

  • Publishing under a pseudonym is not mainly about shielding oneself from offensive remarks; the author's identity is simply not relevant to the content, and a face or LinkedIn link is noise, so the best publishing gives readers the content with the highest signal-to-noise ratio.

    “I think the best way to be successful with publishing online is to have the best signal to noise ratio.”

  • Credentials like a Harvard degree create only limited social capital because they are minted in an industrial process and holders are replaceable; uniquely built social capital, even if less impressive, is more valuable because it cannot be replaced.

    “If you can create social capital in a totally unique way, you're totally irreplaceable, and that is much more valuable.”

  • People's careers follow an S-curve, so the best time to recruit someone is before their first inflection point, when they are still accelerating upward and that progress is visible.

    “People's careers follow an S curve. If you can catch people before that first inflection point, where they're still accelerating upwards”

0.51

Curating a low-noise information diet, not topical news, drives insight and productivity.

7 pointscentrality 3/5
  • For most people it is more important to cut out low-quality information than to seek out good information, because a feed stripped of topical garbage can become very high quality.

    “it's much more important for people to cut out bad information than it is for them to seek out good information.”

  • Information that will not matter within a day is unlikely to yield insight a day later, so consuming topical news wastes attention because the knowledge in one's head goes stale quickly.

    “Anything that's not going to be relevant in a day is unlikely to be a source of solid insight more than a day from now.”

  • Productivity was built by first removing junk food, binge TV, junk information, and unhealthy people, then adding regular exercise and healthy eating, because physical and mental health is upstream of productivity and trying to be productive first is nearly impossible.

    “Your physical and mental health is upstream of your productivity - trying to be productive first is virtually impossible no matter how ambitious you are.”

  • Original ideas require an original information diet, because people who consume the same inputs as everyone else are unlikely to generate unique insights; one should regularly inject random, unique material and high-quality content.

    “in order to achieve some degree of originality, you need to have an original information diet. You're not really that special.”

  • Modern books offer little advantage in information quality over other sources, because distribution and fame produce many weak bestsellers, while certain independent blogs such as Jerry Neumann's and Kevin Simler's are better.

    “there's a lot of really garbage books out there just because somebody has a lot of distribution.”

  • The correct amount of exploration is the question rather than whether to over-explore; many people under-explore, so for them over-exploring may be the correct corrective.

    “Almost by definition, you shouldn't over explore, you should explore the right amount.”

  • Reading daily news leaves people uninformed because the content is out of date; news is only useful at weekly, monthly, or quarterly cadence.

    “You actually are uninformed if you're reading daily news... because it's completely out of date.”

0.47

Popularity is a poor guide to quality, admiration, or durable information in ideas, tweets, and books.

6 pointscentrality 3/5
  • Strong tweets look contrarian but are actually consensus, and people should optimize for saying interesting things rather than true ones, since interesting provocations change people on the margin, and losing followers is healthy.

    “I don't actually try to tweet things that are true because it's almost besides the point.”

  • Any very popular idea that is probably good will be applied in many ways that are bad, because few people are willing to publicly resist widely endorsed sentiments.

    “any very popular idea that's probably good will be applied in a lot of ways that are really bad.”

  • Twitter's success was partly a clown car in a goldmine, so Mayer advises against copying it, and he argues that tweets that signal originality tend to reward followers who are interested rather than those who merely agree.

    “there's this saying in Silicon Valley that Twitter is like a clown car that fell into a goldmine.”

  • Society should admire entrepreneurs rather than pop stars, sports stars, or actors, because entrepreneurs take real risk, and the myth of the successful entrepreneur is important for society even though entrepreneur-porn storytelling is overstated.

    “I think entrepreneurs are better to worship. It's better for society as a whole because entrepreneurs take a lot of risk.”

  • Most podcast guests who give book recommendations probably read only a couple of books a quarter, and books in the modern age are not much higher quality than other sources because popularity and distribution, not quality, drive many of them.

    “there's a lot of really garbage books out there just because somebody has a lot of distribution.”

  • A book still being recommended ten or more years later, or one that is decades or centuries old, is a good signal that it contains durable information worth filling your brain with.

    “if you read something that is still being recommended by people 10 years later.”

0.44

Edge comes from execution and judgment, since ideas and opportunities spread quickly.

5 pointscentrality 3/5
  • Popular technologies like computer vision attract herd-driven attempts at every application, but the right application of a technology is not obvious, so the crowd's enthusiasm does not mean every company should use it.

    “But that doesn't mean that every company should be doing computer vision.”

  • By the time a contrarian idea reaches an ordinary person through reading or online sources, it is no longer contrarian, and because the opportunity vanishes once others discover it, it is probably less likely to work than mainstream approaches.

    “By the time that you hear them, they're not contrarian anymore by definition.”

  • Business strategy knowledge is already widely codified and given away free in shareholder letters and books, so the market for ideas is efficient and the scarce ingredients are execution and the judgment to choose the right strategy.

    “The market for ideas and business strategy is pretty efficient. Berkshire Hathaway and Amazon.com shareholder letters, they basically give it away for free.”

  • Retention cohorts can flatten and then increase, with a long-run cohort buying more coffee at month 18 than at month six, and recent cohorts starting to rise earlier, which Mayer attributes to better scales.

    “that long-run cohort is now buying more coffee than they did in month six.”

  • Once customer acquisition cost on coffee becomes unprofitable, the company can push other products through the existing channel at near-zero acquisition cost, which is why it expects long-term revenue expansion.

    “once our customer acquisition cost blows up on coffee, we can really just start pushing out other products at basically zero acquisition cost”

0.27

Self-directed learning beats expensive schooling and courses for most people.

4 pointscentrality 2/5
  • The batch-processing structure of schooling was rational only before self-education was feasible, because one-to-one teaching was too expensive; the system teaches to the median and loses both the top and bottom students.

    “up until the moment where self-education was really possible, you absolutely need to have the system of education that we have today. Otherwise, it's way too expensive.”

  • People should teach themselves rather than pay thousands of dollars for a course they attend while browsing Facebook, because self-education is now feasible and teachers can serve as helpers.

    “I would just recommend people to actually try to learn things themselves rather than pay somebody $2,000 for a course”

  • Coding is far more accessible than people assume because free, fully documented open-source libraries and abundant examples exist for nearly any task, giving curious people massive leverage without formal training.

    “There are tons of massively powerful libraries and technologies that are totally free to use and fully documented online”

  • A social network built around spaced repetition, where users can save posts into a review schedule, would reward durably relevant content instead of ephemeral one-off posts, changing creators' incentives.

    “you could make a much more meaningful social network if you had some sort of repeatability in the content.”

0.27

Owning the direct customer relationship gives platforms leverage over the suppliers beneath them.

4 pointscentrality 2/5
  • Brands that refuse to sell on Amazon, such as Philz Coffee, show that customer willingness to go out of their way can shift ownership of the relationship away from Amazon, though Amazon also rolls out white-label brands to exploit the default.

    “You can't find them on Amazon. That's part of why.”

  • The entity that owns the direct customer relationship sets the default and controls downstream decisions and money flow, giving it leverage over every supplier beneath it, which is why Amazon captures value even when it does not own the sellers.

    “the person that owns the customer relationship gets to set the web of decision making that happens after that.”

  • Once shipments are bundled, the marginal cost of an extra item approaches zero, so a platform that already delivers multiple boxes weekly could undercut e-commerce incumbents whose prices must cover shipping, as Amazon Prime does.

    “Once you bundle shipments, the marginal cost of an additional item is virtually zero.”

  • Even with ownership of the customer relationship, suppliers retain some power because consumers are brand sensitive, although a platform rotating products like a Pandora-style playlist still gains a large amount of leverage.

    “the person that owns the customer relationship gets to set the web of decision making that happens after that.”

0.27

Problems reframed as information problems can be tackled by technology using the right data and sensors.

4 pointscentrality 2/5
  • If a problem can be reformulated as an information problem and the right data can be gathered, information technology can tackle it, and maturing technologies such as computer vision and cheap smartphone components keep creating openings that incumbents are unlikely to exploit.

    “Fundamentally, if you can reformulate a problem as an information problem, you can tackle it with information technology, if you get the right data.”

  • Although computer vision is a hot trend because the real world has become legible to computers through cameras, it is not the right technology for sensing what is inside people's cabinets or fridges, so weight-based sensors are better suited.

    “I fundamentally don't think computer vision is going to work in people's cabinets or fridges or wherever it may be.”

  • Bottomless is a smart subscription that uses a weight sensor to measure how much coffee remains and how fast it is consumed, so it can ship the next order at the right time rather than on a fixed weekly or monthly schedule.

    “it's essentially a smart subscription that figures out the perfect time to ship you your next order using a weight sensor.”

  • If a problem can be reformulated as an information problem and the right data is gathered, it can be tackled with information technology, and this is the core approach of Bottomless applied to repeatedly purchased items.

    “Fundamentally, if you can reformulate a problem as an information problem, you can tackle it with information technology, if you get the right data.”

0.24

Growth depends on supportive people and compounding habits built on a solid foundation.

3 pointscentrality 2/5
  • Removing people who do not want you to succeed matters because humans are social creatures who internalize others' expectations, so an environment of people who are happy about your success creates an incentive to do better.

    “If the people around you don't want you to thrive, then you actually subconsciously internalize that and don't really want that for yourself, either.”

  • Change happens as momentum flips from a wrong direction to a right one rather than through a single epiphany, and once a solid foundation exists, better habits can be layered on top and compound.

    “Once things are going in the right direction, habits really can compound.”

  • A co-founder who is a woman of action, with focus and follow-through that the speaker lacks, is rare and more valuable than someone who simply has unique insights, and she often steers the company while ensuring the founder carries out his vision.

    “She's the person I would call like a woman of action. Those people are rare in the world. They're very rare. Maybe more rare than people who have unique insights.”

0.24

Technological and market shifts reshape how goods and knowledge are handled.

3 pointscentrality 2/5
  • Spaced repetition, which schedules reviews just before forgetting and spaces them out over days, weeks, and years, can durably store knowledge in memory, making brain implants unnecessary for remembering information.

    “if you use spaced repetition, you can do that already.”

  • Markets cycle from artisanal products to cheap mass production, consolidation that stamps out small producers, a boredom window that triggers artisanal revival at higher prices, and then re-consolidation, as seen in beer and likely in cannabis.

    “then once those people with their huge scale economies have one final victory, once they've completely stamped out all of the small breweries, then suddenly, the landscape for the consumer is just boring.”

  • Making important things legible to computers, such as user locations through smartphones and the physical world through cameras, enabled ride-hailing and Instagram-style businesses, and making stock levels and consumption legible can similarly change how staples are bought.

    “Anytime you can make something important, legible to computers, you can change the world.”

0.20

Equality-focused policies and uniform quality standards tend to backfire and harm societal outcomes.

2 pointscentrality 2/5
  • Equal access to opportunity is good, but the idea gets confused with equal outcomes; policies pursuing equality tend to drop people at the top and bottom, and efforts to lift the bottom above the median indefinitely create constant cost inflation and are impossible to sustain.

    “But it gets confused with equal access to outcome.”

  • Everyone should have basic healthcare, but society's aversion to unequal quality means it does not offer a cheaper, more basic tier of care to everyone, which the speaker thinks would be better than the current outcome.

    “But we have this aversion to lower quality health care for people that we give them for free.”

0.10

Co-founding with a spouse helps early startups through constant communication and personal incentive.

3 pointscentrality 1/5
  • Co-founding with a spouse is an advantage in early startups because critical decisions happen at random moments and need high-bandwidth, constant-uptime communication rather than scheduled meetings during business hours.

    “So it's very high bandwidth because we know each other very well, obviously, and constant uptime on those exchanges.”

  • Working with a spouse creates friction because others assume the arrangement is less serious, though the founders regard their venture as dead serious.

    “There's this tendency for people to think that if you're doing something with your spouse that is less serious.”

  • The kindest act is wanting someone else to do well regardless of how it reflects on oneself, which is rare because most help is self-interested, and having a partner who wanted him to succeed gave him incentive he had lacked.

    “wanting somebody else to do well, regardless of how that reflects on you, is fundamentally kind.”